1. Field of the Invention
This invention relates to name disambiguation, and particularly to systems, methods and computer products for name disambiguation by using private/global directories and communication contexts.
2. Description of Background
Online communication tools, such as e-mails, instant messaging, electronic bulletin boards, and blogs, are commonly used in enterprises for daily works. Even though such communications includes rich information for relationships among persons and knowledge about persons, there are few functions in such tools using results of social network analysis and text analysis. For example, if a name in text is disambiguated automatically, the result will be used for various useful functions in online communication tools, which support/accelerate/enrich online communications.
Some research has been done for a task to identify specific persons appeared in texts, which is known as name disambiguation, name resolution, or name identification. Conventional techniques disambiguate names in a text of a mail body by using relationships expressed in mail archives. At first, person names (ex. “Susan”) in the mail body are recognized, and a set of specific “Susan”s with whom the sender has communicated. And scores for each candidate are calculated based on relationships (such as sender-only or sender+recipients), time scales (daily, weekly, and so on), and so on. Other techniques use k-way spectral clustering method by representing mail archives as a graph structure. Other techniques focus on name variations, such as name abbreviations, name misspellings. Still other techniques use organization information, which uses names in texts for disambiguation of person names. However, the techniques do not disambiguate names by using plural persons included in the body of the same mail, and assume that global mail archives (not a private mail box) are available, which are very difficult to get because of security and privacy matters.